how are politicians informed? witnesses and information
Post on 30-Jun-2022
1 Views
Preview:
TRANSCRIPT
How Are Politicians Informed? Witnessesand Information Provision in Congress∗
Pamela Ban†
University of California, San Diego
Ju Yeon Park‡
University of Essex
Hye Young You§
New York University
Abstract
How are politicians informed and who do politicians seek information from? Therole of information has been at the center for research on legislative organizationsbut there is a lack of systematic empirical work on the information that Congressseeks to acquire and consider. To examine the information flow between Congressand external groups, we construct the most comprehensive dataset to date on 74,082congressional committee hearings and 755,540 witnesses spanning 1960-2018. We showdescriptive patterns of how witness composition varies across time and committee,and how different types of witnesses provide varying levels of analytical information.We develop theoretical expectations for why committees may invite different types ofwitnesses based on committee intent, inter-branch relations, and congressional capacity.Our empirical evidence shows how certain institutional conditions can affect how muchcommittees turn to outsiders for information and from whom they seek information.
∗We thank comments and feedback from Seth Hill, Jaclyn Kaslovsky, Shiro Kuriwaki, Julia Payson, JonRogowski, Jan Stuckatz, and seminar participants at the Institute for Advanced Studies in Toulouse PoliticalEconomy Conference, Korea Advanced Institute of Science and Technology, Washington University in St.Louis, and Yale University.
†Assistant Professor, Department of Political Science. pmban@ucsd.edu.
‡Lecturer, Department of Government. jp20761@essex.ac.uk.
§Assistant Professor, Wilf Family Department of Politics. hyou@nyu.edu.
1 Introduction
How are politicians informed and who do politicians invite to provide information in the
policymaking process? Members of Congress work in complex environments, are time con-
strained, make decisions that have important and potentially far-reaching consequences, and
are constantly pressured to act (Baumgartner and Jones 2015; Curry 2015). In this envi-
ronment, information is one of members’ most important strategic needs and tools as they
consider legislation (Krehbiel 1991). Members may need information about the importance
of problems that they are asked to address (Baumgartner and Leech 1998; Kingdon 1981).
Additionally, members may also require information about the likely impact, effectiveness,
or unintended consequences of policy proposals on their constituents (Krehbiel 1991; Baum-
gartner and Jones 1993) and reelection chances (Hansen 1991; Arnold 1990).
Corporations, think tanks and other groups seek to influence legislators through the
provision of information. Members’ desire for information thus serves as an opportunity for
external groups to enter and gain influence. Providing information as a form of lobbying has
long been characterized in the formal theory literature (Austen-Smith 1993; Lohmann 1995;
Schnakenberg 2017), and is also a key factor in understanding the strategic behaviors that
Congress exhibits when it comes to controlling the bureaucracy and the issues of delegation
(Banks and Weingast 1992; Gailmard and Patty 2012; Turner 2019).
Despite the vast theoretical attention paid to the role of information in legislative organi-
zation and interactions between legislators and external groups, there is a lack of systematic
empirical work on the information that Congress seeks to acquire and consider. Who do
members of Congress seek information from, and how does the content of the information
vary by the identity of information providers? How do institutional conditions such as
divided government and congressional capacity affect information acquisition? Exploring
these questions empirically is essential to understanding the role of information in legislative
institutions and how effectively members enact policies (Volden and Wiseman, 2014).
1
While there are various avenues through which Congress can collect information, com-
mittee hearings and the corresponding witness invitation process present a unique, observ-
able setting that reveals the specific external individuals and information that members of
Congress have selectively sought to consider and convey to other members, interest groups,
the media, and voters during the committee process. We leverage these facts and use witness
testimony to examine the information-seeking behavior of Congress.
In this paper, we present the most comprehensive analysis to date of the information flow
between Congress and external groups by examining the types of witnesses that committees
invite from 1960-2018 and the conditions under which committees turn to some types of
witnesses more than others. First, we introduce our data and describe witness invitation
patterns across 74,082 hearings and 755,540 witnesses who testified in Congress during the 58-
year period of our data. We classify witnesses’ organizational affiliation into 18 types (such
as bureaucrats or labor unions) to capture who Congress invites. We provide descriptive
patterns that track the variation in witness composition across time, by committee, and
by party in the majority. In addition, we show descriptive patterns of how the content of
witness testimonies can vary by their affiliations by using House hearing transcripts for a
subset of our time period.
Given these patterns of how the witnesses who testify in front of committees can vary,
understanding what can influence the invitation patterns of different types of witnesses is
especially of interest. We focus on three categories of explanatory factors which can affect
a committee’s strategic behaviors in inviting external witnesses: committee intent, inter-
branch relations, and congressional capacity. Committee intent—to hold a hearing to explore
a potential legislation or consider a specific bill—captures the committee’s need for partic-
ular types of information. Inter-branch relationships that are determined by whether the
majority party in Congress differs from the party of the president captures the committee’s
incentive in how much information the committee seeks out from executive branch. Congres-
sional capacity such as the size of the supporting congressional organizations captures the
2
committee’s ability to invite external witnesses. We develop theoretical arguments for why
committees may invite different types of witnesses for legislative hearings based on these
three categories of institutional conditions and use our comprehensive dataset to provide
empirical evidence.
First, our results show that committees invite different types of witnesses at different
rates based on committee’s intent: committees turn to think tanks, universities, bureaucrats
at higher rates for hearings without a bill—when committees are using hearings to learn
about an issue area or to produce a potential legislation—and pivot to invite witnesses from
mass-based groups such as labor unions, trade associations, and membership associations at
higher rates for hearings on a specific bill—when committees are using hearings and witness
testimonies to assess the likely impact of the legislation and build a case for the bill under
consideration.
Second, committees’ incentives in managing the inter-branch relationship also has a sig-
nificant impact on witness invitation patterns. We find that during periods of divided gov-
ernment, committees invite relatively lower rates of bureaucrats to testify and instead invite
relatively higher rates of witnesses from think tanks, universities, and from within Congress
itself. Furthermore, we find that this result is particularly pronounced when hearings are
held on issues that the president prioritizes, compared to when hearings are held on issues
that the president does not prioritize. These results are substantively important especially
considering how the existing literature has characterized bureaucrats’ advantages in informa-
tion and expertise on policy implementation vis-a-vis Congress (Gailmard and Patty, 2012).
Our findings provide evidence for how committees limit the amount of expert information
from an executive branch favorable to the opposing party’s president and, instead, open a
door to external groups such as think tanks and university researchers to compensate for the
relative loss of information from bureaucrats (Banks and Weingast, 1992).
Third, to examine how congressional capacity influences witness invitation patterns, we
examine how the 1995 reform by a new Republican majority in the House, which downsized
3
the internal resources of Congress, affected the information acquisition behavior of commit-
tees. Using a difference-in-differences design, we show how the elimination of the Office of
Technology Assessment (OTA)—a supporting agency which provided advice to Congress on
emerging technologies and other scientific matters—drove committees to change their behav-
ior in how much and from whom they seek external information. We find that committees
who relied most on internally-produced information within Congress suffered a drastic drop
in the number of technical and scientific witnesses they could manage to invite in the wake
of the OTA elimination.
Broadly, this article makes three notable contributions. We construct the most compre-
hensive database to date on congressional committee hearings and witnesses who appear
before Congress; our data not only greatly expands the year coverage of hearings and wit-
nesses, but also provides valuable data such as the individual affiliations and types of these
witnesses. In addition, while there has been ample theoretical attention devoted to the role
of information in legislative organization and behavior, our findings fill a gap by providing
empirical evidence on how institutional conditions can affect how much legislatures turn to
outsiders for information and who in particular they turn to. Lastly, and more generally, this
paper pushes forward our understanding of how external groups seek to influence legislators
through the provision of information at congressional hearings. By documenting which ex-
ternal groups get invited and whether the type of information varies by group affiliations,
our research highlights the potential role of external groups in shaping legislative processes.
The next section provides a primer on congressional hearings and witnesses, followed
by introducing our new dataset on witnesses for the period 1960-2018 and presenting key
descriptive statistics on the witenss invitation patterns and the variation in the type of in-
formation they provide. We then present theoretical arguments for how certain institutional
conditions—committee intent, inter-branch relations, and congressional capacity—affect who
Congress decides to invite as witnesses and provide empirical evidene for our theoretical ex-
4
pectations. The final section discusses the implications of the findings and suggest future
work.
2 A Primer on Congressional Hearings and Witnesses
The committee stage is a prime market for information. The importance of hearings during
the committee stage has been noted by the congressional literature (Oleszek, 1989; Deering
and Smith, 1997), and has been the setting of previous studies on communication and infor-
mation flow among legislators, interest groups, and bureaucrats (e.g. Leyden 1995; McGrath
2013). Previous research and case studies have shown how legislative outcomes and the con-
tent of bills have been affected by the information that is aired and discussed at committee
hearings (Burstein, 1999), and by conflicts among witnesses’ testimonies about issue framing
during committee hearings (Baumgartner and Jones, 1993). Furthermore, interest groups
have been shown to be particularly interested in providing information at the committee
stage (Leyden, 1995).
Congressional committees hold these hearings to carry out their work. Namely, commit-
tee hearings are held for one of four purposes: (1) to collect information and opinions on
legislation, (2) to conduct oversight on executive agencies, (3) to investigate events, and, in
Senate committees, (4) to consider presidential nominations as part of confirmation processes
(Heitshusen 2017). In any type of committee hearing, members from both the majority and
minority parties are given the chance to make statements, ask questions, debate opinions,
invite outside witnesses to testify, and question outside witnesses about the topics at hand.
In general, hearings provide an opportunity for committee members to engage with external
witnesses as members collect information, discuss ideas, and formulate policy. Witnesses who
appear in Congress only appear in front of congressional committees; there are no witnesses
who testify on the floor.
5
Members during the committee stage are thus faced with the decision of who – which
witnesses – to invite to testify and provide information. Committee members, with their
committee staff, will identify potential witnesses for a hearing (Heitshusen, 2017; Davis,
2015). There is no limit to the number of witnesses that may be invited.1 During the
consideration of potential witnesses, the committee members of the majority party may
weigh in on the selection of witnesses and provide recommendations to the chair, though
the chair possesses the gatekeeping power over which witnesses ultimately get invited to
testify. Since 1970, the minority party’s committee members have been granted protection
by chamber rules to call their own witnesses of choice on at least one day of each hearing.
In some cases, witnesses are selected to represent various reasonable points of view; in
other cases, witnesses are selected to represent a specific point of view (Heitshusen, 2017;
Davis, 2015). When choosing witnesses, committees are faced with making various choices,
such as how many witnesses to invite, or what types of witnesses to invite. When thinking
about what types of witnesses to invite, witnesses can vary by numerous characteristics,
such as gender, ideological leaning, expertise in the issue area, etc. While there can be an
unending list of characteristics that can describe witnesses, many salient characteristics may
not be known for certain or available to a committee when they are inviting witnesses, such
as precise knowledge of a witness’ ideology.2 However, one clear, salient, and easily accessible
characteristic for committees to use is a witness’ organizational affiliation. In the existing
literature, organizational affiliations (e.g. business interests or membership organizations)
have been used to characterize groups present and active in the political process (Yackee
and Yackee 2006; Schlozman et al. 2015). Although there is variation in the resources and
1Witnesses who receive invitations are often eager to testify, but if not, committees can exercise theircongressional subpoena power to compel a specific witness to testify (Davis, 2015).
2The ideology of external groups has received vast theoretical attention in the literature of legislative or-ganization and lobbying (e.g., Kollman 1997). While witness ideology may be of interest to scholars, theideology of witnesses is difficult to determine accurately and systematically across our extensive dataset.Although the ideologies of witnesses could be extracted by using data based on campaign contributions(Bonica, 2016), not all witnesses or witnesses’ organizational affiliations have made political donations thatwould be necessary to be ideologically scored. This limitation will result in significant missing data issuesif we focus on the ideology as a key characteristic of witnesses.
6
opinions within the same affiliation type, affiliation types can be a good proxy for the overall
composition and diversity of the invited witnesses from the perspective of the committees.
Thus, while the process for inviting witnesses is rather straightforward, there can be
a variety of factors that can affect which witnesses – from which types of affiliations – are
ultimately invited to testify and appear before committees, which we expand upon in Section
4. In the next section, we describe our comprehensive dataset and start with descriptive
patterns to illustrate what witness compositions in Congress have looked like.
3 New Data on Congressional Hearings and Witnesses
We constructed a new dataset on congressional committee hearings and witnesses from 1960
to 2018. This data was collected from ProQuest Congressional. The dataset includes full
names of the 755,540 witnesses who appeared in 74,082 hearings of the House, Senate and
Joint standing committee hearings during this period and their organizational affiliations.
For each hearing, we extracted the following hearing-level information: title, date, the name
of the committee that held a hearing, summary of hearing contents, and any bill numbers
considered in each hearing.
Compared to the existing data on congressional hearings used by scholars, our database
will be the most comprehensive in terms of both the year coverage and the breadth of
information.3 Although some extant literature has analyzed witnesses who testified in a small
selection of hearings in a limited period of time (e.g., Leyden 1995; Flemming, MacLeod,
and Talbert 1998), the congressional scholarship has never systematically built a complete,
extensive dataset on witnesses who testified in committee hearings.
Based on the raw data we have collected, we further processed the data by construct-
ing key variables that capture witnesses’ characteristics. Our key interest is the witnesses’
affiliations. As stated previously, affiliations have been used to characterize groups in the
3For example, the data on congressional hearings as part of the Policy Agendas Project (PAP) start from 1970and do not provide any information about witnesses. See more at https://www.comparativeagendas.net/
7
political process, and other characteristics such as ideology or expertise on issues are either
difficult to measure or unavailable for an extensive set of witnesses. Therefore, we focus on
the affiliation of witnesses, which provide a good approximation for the types of external
groups that are invited to congressional hearings. We classified witnesses’ affiliations into
the 18 types. Table 1 presents the 18 types, percentage of each type in our dataset, one
example of a witness affiliation (or title) in each type, and the 9 broader parent categories
of the 18 different types that are used for graphical presentation of our data later.
This classification was a careful procedure: a) first, we constructed a list of affiliations
of potential witnesses based on existing data from five different sources which we explain
in more detail in the next paragraph, then b) assigned one of our predetermined categories
to each organization or job category, and finally c) merged the list to our new dataset on
witnesses by matching the affiliations from both sides of the data. This process involved both
automated match and extensive manual cleaning. It results in a dataset that, for the first
time, systematically catalogs the organizational affiliation of every witness who has testified
in Congress from 1960-2018.4
There are five sources from which we retrieved the list of organizations, groups and federal
bureaucratic agencies to use in the above procedure. First, we extracted names of clients and
lobbying firms from the Lobbying Disclosure Act (LDA) data available at LobbyView.org
(Kim 2018). Second, we retrieved a list of organizations or employers of political donors
from the Database on Ideology, Money in Politics, and Elections by (Bonica 2016). Third,
we collected a list of departments and agencies of the federal bureaucracy from the Office
of Public Management (OPM). Fourth, we also utilize the Washington Representatives Di-
rectory which includes organizations that are active in Washington DC politics. Lastly, we
collected a list of foreign governments from the Correlates of War Project. Together, these
4There are 23,519 out of 755,540 witnesses (3.1%) who have missing affiliation information. These cases arewhen the witness information only includes names of witnesses without further information. There is nosystematic patterns of missingness in the affiliation type by year or committee.
8
Table 1 – Types of Witness Affiliation
Type Composition (%) Example CategoryAgriculture 1.64 American Farm Bureau BusinessCorporation 8.85 Ford Motor Co. BusinessTrade Association 6.48 Chamber of Commerce BusinessBureaucrat 24.98 Department of Defense BureaucratCongressional 8.85 Congressional Budget Office CongressionalState&Local Government 10.56 Mayor Local Gov(K-12) Educational 1.06 Superintendent Local GovThink Tank&University 8.45 MIT ResearchMembership Association 9.44 Veterans of Foreign Wars Membership Assoc.Non profit 7.52 Environmental Defense Fund NonprofitLabor Union 2.29 AFL-CIO LaborJudicial 0.94 District Court OtherLawyers&Lobbyists 1.33 American Bar Association OtherHealthcare 1.66 American Hospital Association OtherNative American 1.24 National Congress of American Indians OtherReligious 0.60 US Catholic Conference OtherCitizen 2.77 Resident OtherInternational 0.39 World Bank Other
Total Number of Witness 732,021
five datasets identified 1,063,223 unique names of the groups with which witnesses can be
potentially affiliated.
In addition, we constructed committee-level variables, explained in a later section, and
merged them to our dataset on witnesses. Next, we classified hearings into three types:
legislative, oversight or investigative, and nomination hearings.5 Lastly, we merged issue
areas of each hearing from the Policy Agendas Project database on congressional hearings.
3.1 Descriptive Statistics: Witness Compositions
Our new dataset shows that the number of witnesses who appear in Congress varies signif-
icantly over time. Figure 1 illustrates the total number of witnesses who have appeared in
5We identify nomination hearings as hearings that considered a nomination. For oversight or investigativehearings, we follow McGrath (2013) and classify non-nomination hearings as oversight or investigative ifthe PAP’s description of that hearing contain one or more of the following words: “oversight,” “review,”“report,” “budget request,” “control,” “impact,” “information,” “investigation,” “request,” “explanation,”“president,” “administration,” “contract,” “consultation,” “examination.” This is the same set of wordsused to filter for these types of hearings by McGrath (2013). Finally, we classify hearings that are notoversight or investigative, nor nomination hearings, as legislative hearings.
9
Figure 1 – Number of Hearings and Witnesses in Congress Over Time
0100200300400500600700800900
1000110012001300140015001600170018001900200021002200230024002500
1960 1970 1980 1990 2000 2010Year
Num
ber
of H
earin
gs
HouseSenate
Hearings
0
2500
5000
7500
10000
12500
15000
17500
20000
22500
25000
27500
30000
1960 1970 1980 1990 2000 2010Year
Num
ber
of W
itnes
ses
HouseSenate
Witnesses
Notes: The left figure shows the total number of hearings held by congressional committees in each
two-year Congress in each chamber. The right figure shows the total number of witnesses who
have appeared in committee hearings in each two-year Congress in each chamber. Each Congress is
plotted by its first year.
each two-year Congress in each chamber from 1960 through 2018, as well as the number of
hearings held by committees in each two-year Congress. A couple of main patterns emerge.
First, the peak in the number of witnesses occurred in the 1970s, where the number of ex-
ternal witnesses topped out at 29,665 in the 95th House (1977 through 1979) and 17,027 in
the 93rd Senate (1973 through 1975). This is likely in accordance with the increase in the
number of subcommittees that resulted from the Subcommittee Bill of Rights in 1974; an in-
crease in the number of subcommittees likely increases the number of hearings held and thus
the number of witnesses. These maximums then decrease across time until the minimums
seen in most recent years; the number of witnesses in Congress experienced a decline since
the 1980s, with around five times fewer witnesses testifying in Congress now than at the
peak in the late 1970s.6 One possible contributor to this is a reform in 1995 that drastically
6While the trend in the number of witnesses does sharply decrease across time, the points seen in the lasttwo years of the graph (2017-2018) do not include all hearings held, as hearings are still not completelymade available for the most recent Congresses. For instance, classified hearings that happened in recentCongresses may not yet be declassified (compared to classified hearings that have been declassified acrosstime).
10
cut the number of subcommittees, which had the opposite effect as the 1974 reform; cutting
subcommittees means fewer chances for subcommittee hearings and thus witnesses (Deering
and Smith, 1997).
Two other overall patterns between the two chambers can be seen from Figure 1. First,
the number of witnesses follows similar trends in the House and the Senate; when the number
of witnesses rises [falls] in one chamber, the number of witnesses rises [falls] as well in the
other chamber. Second, the number of witnesses in the House for any given year has always
been greater than the number of witnesses in the Senate. Finally, while Figure 1 presents
the total number of witnesses in each chamber, Figures A1 and A2 in the Appendix present
the number of witnesses who have appeared by committee over time.
Figure 2 – Witness Affiliations Over Time
0%
25%
50%
75%
100%
1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015
Year
Com
posi
tion
BureaucratBusinessMembership AssocResearchCongressionalLocal GovNonprofitLaborOther
The composition of witness affiliations across time is presented in Figure 2. For illus-
trative purposes, we grouped the 18 affiliation types we identified through the procedure
described previously into 9 parent categories for Figure 2.7 On the whole, bureaucrats rep-
resent the plurality of witnesses at any point in time. Over time, there has a been a gradual
7Appendix Figures A6 and A7 present trends in the number of witnesses by specific type across time, forthe House and the Senate, respectively.
11
increase in the percentage of witnesses from the think tank and research category, and a de-
crease in the percentage of witnesses from membership associations and local governments.
Figure 3 – Witness Affiliations By House Committee
0%
25%
50%
75%
100%
Armed
Ser
vices
Overs
ight a
nd R
eform
Fore
ign A
ffairs
Scienc
e, S
pace
, and
Tech
nolog
y
Appro
priat
ions
Vete
rans
' Affa
irs
Financ
ial S
ervic
es
Trans
porta
tion
and
Infra
struc
ture
Energ
y and
Com
mer
ce
Small
Bus
iness
Budge
t
Judic
iary
Natur
al Res
ourc
es
Way
s and
Mea
ns
Agricu
lture
House
Adm
inistr
ation
Educa
tion
and
Labo
r
Ethics
Rules
Committee
Com
posi
tion
BureaucratBusinessMembership AssocResearchCongressionalLocal GovNonprofitLaborOther
House
In addition to these trends across time, interesting variations appear when looking at
committee-by-committee descriptive patterns. For instance, Figure 3 shows the average wit-
ness affiliations by committee in the House.8 Immediately, it is clear that committees can
differ widely by the type of witnesses they favor. Bureaucrats strongly dominate the pres-
ence of witnesses in the Committees on Armed Services, Foreign Affairs, Veterans’ Affairs,
and Government Operations; this is perhaps due to the high administrative focus of these
committees. On the other hand, business witnesses command relatively more presence in
the Agriculture, Banking, Energy and Commerce, and Small Business Committees, reflect-
ing the tendency of these committees to request information from external sources in these
industries.
8Figure A3 in the Appendix shows the equivalent for Senate committees.
12
3.2 Variation in the Content of Witness Testimonies
These descriptive patterns show how the composition of witnesses, in terms of their affilia-
tions, has varied. While witness affiliations may be the clearest, most relevant characteristic
of a witness present to committees when they choose witnesses, do affiliations capture mean-
ingful differences in information? In this section, we illustrate one way in how the content
of witness testimonies can vary by their affiliation.
The content of witness testimonies can vary in numerous ways; one measure of infor-
mation that the existing empirical literature has focused on is the amount of falsifiable
statements about the policy under consideration. Esterling (2004, 2007) terms this type
of information analytical discourse, while other scholars have termed this type of informa-
tion as “policy-analytic knowledge” or “technical information” (Bradley, 1980). This stands
in contrast to non-analytical information, for example conveyed in the form of anecdotes
or personal information, which other scholars have categorized as “ordinary knowledge” or
“experiential discourse” (Esterling, 2007). While non-analytical information is also useful
politically, especially for politicians to be able to understand and connect with constituents
(Esterling, 2007), it is analytical information that is the necessary input to technical policy
development and is the type of information that positive theories have mostly focused on
(Krehbiel, 1991). Further, recent scholarly discussion on the declining analytical capacity of
Congress adds additional importance to understanding the quantity and quality of analytical
information provided by external witnesses (Burgat and Hunt, 2020). Following this, for the
purposes of descriptive statistics in this section, we look at the amount of analytical infor-
mation present in witness testimonies as a descriptive example of how witness testimonies
can vary according to their affiliation type.
To do so, we collected hearing transcripts for the House from the 105th through 114th
Congresses from the Government Publishing Office and parsed the transcripts by each state-
ment or speaking instance (including speeches, questions, answers, and other declarations)
13
made by witnesses.9 In order to measure which types of witnesses tend to provide more
analytical information in hearings, we quantify three aspects of witnesses’ testimonies: How
many words each witness spoke in a hearing; how many keywords which may convey analyt-
ical information that each witness used in a hearing; the proportion of these keywords out of
all the words that each spoke in a hearing. We take the proportion of keywords as the main
variable of interest as it best shows how efficiently a witness conveys analytical information
in their testimony. We identify the set of keywords that may contain analytical information
in three ways: words related to cognitive orientation from the Harvard IV-4 dictionary, words
frequently appearing in information-seeking statements as identified in Park (2021), and any
additional word stems that are similar to those in the first two groups. Details on how we
identify keywords through these approaches can be found in Appendix Section B.
As we are interested in how witness affiliations correlate with the amount of analytical
information present in the witness testimonies, the main independent variables of interest
are witness affiliation types. Figure 4 presents the coefficients on witness type fixed effects,
from an ordinary least squares regression that predicts the proportion of keywords that a
witness uses in a hearing. We include hearing- and committee- level controls, along with
issue, committee and congress fixed effects; the full regression model and results are presented
in Appendix Table A1. The reference group is set as the witnesses representing nonprofit
organizations.
The figure shows that bureaucrats and witnesses from think tanks and research insti-
tutions tend to give testimonies with the highest proportion of analytical information. On
the other hand, individual citizens without an organizational affiliation and those repre-
senting religious institutions tend to provide the lowest proportion of analytical information,
9Based on the committee membership assignment data by Stewart and Woon (Stewart III and Woon 2017),committee members’ statements and speaking instances in the transcripts are identified by their last names.Similarly, witnesses are identified by their last names based on the witness data we have. Then, we useonly the witnesses’ testimonies for this study.
14
Figure 4 – Proportion of Keywords by Witness Type
Notes: Vertical lines indicate 95% confidence interval. The reference group is the witnesses from
nonprofit organizations.
which seems naturally consistent and lends confidence that our measurement is substantively
valid.10
There is a clear gap between the types of witnesses who provide the most and least
analytical testimonies. Based on Figure 4, the difference between the coefficients for the
bureaucrats and citizens is 0.017. Given that the witnesses in this analysis tend to speak
about 1,923 words in a hearing on average, bureaucrats are likely to use 32 more analytical
keywords in a hearing on average than ordinary citizens. To further examine whether this dif-
ference is noticeable substantively, in Table 2 we provide two sample statements that include
10Alternatively, when we look at the second model in which the dependent variable is the number of keywordsspoken, the top two and bottom two groups remain the same. The coefficient plot for this model is presentedin Figure A8 in the Appendix.
15
Table 2 – Examples of the Most and Least Analytical Testimony
With the largest proportion of keywords With the smallest proportion of keywordsStatement “When projects are authorized, when there
is a Chief’s Report and the Congress au-thorizes a project, the economic analysisthat is done on that calculates a benefit tocost ratio. And that benefit to cost ratio isbased on a 3.125 discount rate. When theOffice of Management and Budget evalu-ates projects for funding, including in thePresident’s budget, that benefit to cost ra-tio is evaluated at a 7-percent discountrate. So the budgeting discount rate is dif-ferent from the authorization discount ratethat’s used.”
“When Michael came home that night andI confronted him and was talking to him,he had eye contact like we do now. Butwhen he was sitting on the sofa and nobodywas confronting him, he was comatose. Hewas in the ozone. He was sitting with hismouth hanging open, staring at the floor. Iknew that there was something wrong withhim that night. I could tell that he hadtaken something.”
Speaker Jo-Ellen Darcy, Assistant Secretary, CivilWorks , Department of Army
Brad Alumbaugh, A parent of drug over-dose victim
Type Bureaucrat Citizen
at least 50 words. The first one is the statement with the largest proportion of keywords
among those made by bureaucrats and the second one is the statement with the smallest
proportion of keywords among those made by citizen witnesses. In the first statement, a
bureaucrat witness informs the committee members about different discount rates applied
at different stages of policy-making and implementation process. On the other hand, in the
second statement, a parent of a drug overdose victim explains his anecdotal experience with
his son. The contrast between these two statements shows that the difference in the level of
analytical information is successfully captured by our measurement. We provide additional,
more representative sample statements to confirm this finding in the Appendix Section B,
as well as details of a statistical validation with human coders.
The pattern demonstrated in this section shows that not all witness testimonies are the
same in the type of information they provide. When examining the relative amount of ana-
lytical statements present in testimonies at hearings, it is clear that committees may receive
different amounts of analytical information based on what types of witnesses they invite.
This motivates how the composition of witness invitations hold important implications for
committees, as not only do witness invitations show who committees select to hear from, but
16
they also signify the different types of information that committees may ultimately receive
from witnesses.
4 How Institutional Factors Affect Witness Invitations
The descriptive patterns in the previous section provide a picture of how the witnesses
who testify in front of committees can vary. As we are interested in who Congress invites
to provide information to produce policy, we focus on legislative hearings. What affects
Congress’ decision of who to invite to testify and provide information in legislative hearings,
and under what conditions can we expect committees to invite more or less of certain types
of witnesses? In this section, we present a theoretical framework that incorporates how
three categories of explanatory factors can affect the choices of who committees turn to
for external information. We characterize the strategic decision of a committee’s witness
selection as a function of the committee’s intent for the hearing, how inter-branch relations
politicize the information supply from within the federal government, and the extent to
which the internal resources of Congress enable the selection and arrangement of witnesses.
This spans three factors that can affect a committee’s need (committee intent), incentives
(inter-branch relations), and abilities (congressional capacity) in inviting external witnesses.
This framework generates three categories of testable predictions regarding how institutional
factors affect the composition of witnesses chosen by committees.
Committee Intent. We start with the relationship between committee intent and types
of witnesses. As explained in Section 3, committees may hold legislative hearings at any point
in the policy-making process for an issue. A hearing can be held on a legislative issue without
any specific bills under consideration, or a hearing can be held to discuss a specific bill that
was introduced and is under consideration. A committee’s information-seeking behavior may
thus vary based on whether a hearing is to develop a potential bill or if a hearing is on a
17
specific bill.11 A legislative hearing that is on a topic for potential legislation but does not yet
have a specific bill under consideration likely reflects a committee’s intent to hold a hearing
primarily for learning information and may seek information to learn about the issue area
or potential legislation. Alternatively, in a hearing on a specific bill, since a committee has
already decided to hold a hearing on a specific bill, it is thus more likely to have the intent
of conveying a specific view, message, or justification for the bill in the hearing. While the
committee may still have various intentions in a hearing, we focus on the relative intention. A
hearing on a specific bill is likely to be relatively more about strategically using the questions
and answers between members and witnesses to inform observers of the hearing (e.g. interest
groups, media, lobbyists, other members) of any positions or reasonings for or against the
bill, compared to a hearing without a bill attached.
Committees may wish to seek different types of witnesses based on their intent for the
hearing; they may change the scope of information they seek and the sources that they
invite to testify. When a committee is seeking witness testimony to learn about an issue
area or potential legislation, they may wish to seek expert information about the details
of what is needed to create policy (i.e. from a narrower set of witnesses that can provide
expert information). On the other hand, when a committee is seeking witness testimony to
convey a specific view or message about the bill, they may wish to seek information from a
wider variety of witnesses, such as groups that are likely to be affected (either positively or
negatively) by the legislation in order to message about the bill or build a case for the bill
under consideration. This leads to our first hypothesis:
Committee Intent Hypothesis: Committees will invite a narrower set of witnesses and rela-
tively more witnesses who can provide expert information in legislative hearings without a
11A committee’s intent when the committee schedules a legislative hearing is unobservable directly. This isbecause there is no systematically available information at the time a hearing is scheduled that directlyshows what the committee’s intent is; this intent is “private” information. Any content of the hearingalso cannot be used to gauge the committee intent as that would be post-treatment. To gauge committeeintent, then, we use a variable that reflects what a legislative hearing’s main purpose is – whether a hearingis on a bill or not.
18
bill attached compared to legislative hearings on a specific bill. Committees will invite more
varieties of witnesses and relatively more witnesses from groups that are likely to be affected
by legislation in legislative hearings on a specific bill compared to legislative hearings without
a bill attached.
Inter-Branch Relations. Second, we consider whether Congress searches for more
(or less) information, and from different sources, when there is divided government versus
unified government. Divided government creates issues for legislative control over the im-
plementation process and, thus, Congress has created numerous legislative and procedural
solutions to increase its influence on behaviors of executive branch: for one, they can design
agencies to be more insulated from the president’s influence (Lewis, 2003), or they can write
more detailed laws (Huber, Shipan, and Pfahler, 2001) to reduce the discretion delegated to
the bureaucracy (Epstein and O’Halloran, 1999).
Congressional hearings are another tool that the legislative majority can employ to exer-
cise control over the executive branch. For example, scholars document that divided govern-
ment is strongly related to Congress’ use of investigative hearings on the executive branch’s
conduct (Kriner and Schickler, 2016). While Kriner and Schickler (2016) examine investiga-
tive hearings in particular, the logic of attempting to manage the power of the executive
branch through hearings can be applied to legislative hearings as well. Bureaucrats as wit-
nesses are an important group to consider, as they have been characterized as possessing an
informational advantage in policy production and implementation over Congress (Gailmard
and Patty 2012; Patty and Turner 2021), and our own descriptive patterns in Section 3.2
reveal that bureaucrats provide the relative highest levels of analytical information. As seen
in the descriptive patterns in Figures 2 and 3, bureaucrats indeed comprise a substantial
number of the witnesses that committees call to testify. While there are many career civil
servants in the bureaucracy, the president is the head of the executive branch and, addition-
ally, names political appointees who oversee and directly manage career bureaucrats.
19
The informational advantage and policy expertise that bureaucrats possess – but also the
connection between bureaucrats and the heads of the executive branch – raise a strategic
question for committee chairs as they consider whether to invite bureaucrats as witnesses
for legislative hearings. This becomes especially salient when there are policy disagreements
between the legislative and executive branches of the government (and perhaps in particular
on issues that the president prioritizes), which is more likely during periods of divided gov-
ernment. Thus, when the majority party in Congress differs from the party of the president,
committees are faced with the potential of bureaucratic witnesses representing the opposing
party (and bureaucratic officials are faced with the potential of providing valuable informa-
tion to a Congress controlled by the opposing party). As a result, committees may be more
likely to turn to other sources of information, such as other types of witnesses or internal
congressional sources. This leads to our second hypothesis:
Inter-Branch Relations Hypothesis: Committees will invite relatively fewer bureaucrats as
witnesses in legislative hearings during periods of divided government compared to periods
of unified government.
Congressional Capacity. Third, we consider how internal capacity of the Congress
affects the witness invitation patterns. Scholars describe congressional capacity as the level
of internal resources of Congress, with one main internal resource being internal congres-
sional support agencies. The Congressional Budget Office, Congressional Research Service,
Government Accountability Office, and the former Office of Technology Assessment (OTA)
make up the set of internal support agencies that were created to serve and assist members
and committees in their workflow (Kosar, 2020). In general, these internal support agencies
provide information to Congress that help identify matters that Congress should address
and attend to, arm legislators with specialized information, and help rebalance intra-branch
information asymmetries (Baumgartner and Jones, 2015).
20
This form of congressional capacity received a shock in 1995, when the Republicans
became the House majority party for the first time since 1952. One of the core agendas
of House Speaker Newt Gingrich’s “Contract with America” platform was to downsize the
government and the legislative branch was not immune to changes. The Republican majority
in the House eliminated funding for the Office of Technology Assessment (OTA) and cut
resources for the other internal congressional support agencies as part of their 1995 reform.
Congress had created the OTA in 1972 to study emerging technologies and to provide
advice to Congress on these technologies and other scientific matters. The information
from the OTA, and other internal support agencies, were often routed through congressional
committees – an individual member of Congress could not request a study or report from the
OTA, but a congressional committee could. As a result, committees that were particularly in
need of scientific and technical advance frequently requested information from the OTA, and
the OTA acted as a provider of information and a source of expert staffers internally within
Congress. Committees who relied on the OTA reported not just the benefit of internal
information from the OTA, but also of trusted relationships with OTA staff that helped
committees navigate scientific research and sort through the amount of available expertise
and competing expert opinions (Tudor and Warner, 2019; Johnson, 2019). Thus, with the
elimination of the OTA in 1995, committees who frequently relied on the OTA suffered an
immediate cut in internal information and the absence of a group of OTA staffers who liaised
between committees and the scientific community.
We examine how the elimination of the OTA in 1995 affected the invitation patterns
of external witnesses for committees who had depended on the OTA for information and
expertise. On one hand, with the defunding of the OTA, committees that had relied heavily
on internal sources of information may increase their efforts in inviting external witnesses,
especially research-based witnesses who can provide technical and analytical information,
in order to compensate for the loss of internal information that had been provided from
21
the OTA. This leads to the hypotheses that emphasizes the substitution effect for the third
factor:
Congressional Capacity Hypothesis 3A: Committees that relied more on the OTA will invite
relatively more witnesses from think tanks and research organizations in legislative hearings
after the elimination of the OTA.
On the other hand, without the OTA’s advice and guidance, those committees may have
a reduced capacity to even identify or facilitate the invitation of scientific witnesses on their
own. The process of witness selection takes time and resources, especially for the types of
witnesses that require relatively more effort to identify, research, and prepare. What’s more,
the 1995 reform also drastically cut committee staff across all committees.12 As staffers are
integral to arranging witnesses for hearings, sufficient numbers of committee staff may need
to be maintained in order to support a committee’s search for external information. The
elimination of the OTA, along with a substantial cut in committee staff, could result in a
more drastic reduction of expert witnesses in the committees that relied more on the OTA,
even though demand for those types of witnesses may have increased. This leads to the
hypothesis that emphasizes the amplifying effect of the loss of congressional capacity:
Congressional Capacity Hypothesis 3B: Committees that relied more on the OTA will invite
relatively fewer witnesses from think tanks and research organizations in legislative hearings
after the elimination of the OTA.
5 Empirical Evidence
In this section, we provide empirical evidence for the theoretical intuitions on how institu-
tional conditions affect the patterns of witness invitations. To do so, we focus on legislative
hearings in the House of Representatives.
12Figure A10, which presents the patterns of committee staffing in each standing committee in the Houseacross time, shows that there were sharp declines in the number of committee staffers across the board.
22
5.1 Committee Intent and Witness Invitations
We investigate the effect of committee intent on witness invitation patterns by examining
how the quantity of witnesses and composition of witnesses at a legislative hearing vary based
on whether the committee intends to use the hearing relatively more to learn information
about an issue area or relatively more to convey information to produce a message or justify
a position on an issue, as previously discussed.
We use the following regression and ordinary least squares estimation:
Yhict = βHearing Characteristicshict + γCommittee Characteristicsct + αi + αc + αt + εhict
where the subscripts indicate hearing h, committee c, issue i, and congress t.13 The outcome
variable Yhijt will measure (1) the quantity of witnesses and (2) the diversity of witness types
present at a given hearing, along with the percentage of witnesses from each affiliation type
present at the hearing. Hearing Characteristics contain the main hearing-level variable of
interest that proxies the committee’s intent in the hearing: whether the hearing had a bill
attached to it. Besides this key explanatory variable, we also include control variables such as
Subcommittee (which equals 1 if the hearing was held at the subcommittee level, and equals
0 otherwise). We include fixed effects by committee, issue, and congress. While we use a
committee level fixed effect, we also include committee-level control variables in Committee
Characteristics ijt such as the total number of committee members and the absolute difference
in DW-NOMINATE scores between the committee chair and the floor media, as they may be
of interest in the estimated results.14 Standard errors are clustered at the committee level.
13The issues j represent the 21 major topics from the Policy Agendas Project.
14Additional committee-level time-varying controls are the absolute difference in the DW-NOMINATEscore between the Democrats and Republicans in the committee, and the absolute difference in the DW-NOMINATE score between the committee median and floor median.
23
Figure 5 presents the coefficient plots for the selected outcome variables of interests when
a hearing considers a specific bill.15 The outcome variable “No.Witness” is the number of
witnesses invited to testify at the hearing. When hearings consider a specific bill, committees
tend to invite more witnesses. The outcome variable “Diversity” represents the diversity of
witness types and is based on the Herfindahl index of the witness types that are present in
a given hearing: for the eighteen possible witness types, we calculate each type’s share of
the total number of witnesses in a given hearing and sum the squares of these shares. For
ease of interpretation, we take 1 minus this Herfindahl index in order to create our outcome
variable, such that a higher value will indicate more diversity in witness types in a hearing,
and a lower value indicates less diversity. The results in Figure 5 show that hearings which
consider bills tend to invite more witnesses, and a higher diversity of witnesses, compared
to hearings that are held without specific bills attached.
Which types of witnesses are invited more often during hearings without a bill compared
to during hearings with a bill when there is already a specific bill developed? These types of
witnesses are the ones with a negative significant coefficient in Figure 5. First, the share of
one type of witness, bureaucrats, is of interest due to their particular knowledge about policy
production and needs (Bendor, Taylor, and Gaalen 1987; Gailmard and Patty 2012; Patty
and Turner 2021) and higher levels of analytical information (as we illustrate in Section
3.2) that they can bring to committees. While we will investigate how partisan control of
government affects the presence of bureaucrats in committee hearings in the next section,
here, we show that the relative frequency of bureaucrats is related to the committee’s intent
in hearings. The results show that committees tend to seek out bureaucrats – their analytical
information and expert information about policy production and needs – more often when
committees are not considering a specific bill, compared to when committees already have a
specific bill developed.
15Table A2 in the Appendix presents the results that investigates the effects of hearing characteristics onwitnesses.
24
Figure 5 – The Effect of Hearing Considering Specific Bills on Witness Invitations
−10.00
−5.00
0.00
5.00
10.00
No.
Witn
ess
Diver
sity
Burea
ucra
t
Res
earc
h
Cor
pora
tion
Labo
r
Trade
Ass
oc.
Mem
bers
hip
Assoc
.
Notes: Each plot indicates the regression coefficient for each outcome measure (x-axis). Y-axis shows
the regression coefficients; “No. Witness” is the number of witnesses, “Diversity” is the Herfindahl
index, and the others are the percentage shares of witnesses. The groups not shown in the plot
have coefficients that are not statistically significant. The plots are presented with 95% confidence
interval.
Second, the results also show that committees invite relatively more witnesses from think
tanks or universities (“Research”) for hearings without a specific bill attached compared to
hearings on a specific bill. Think tanks and universities are of interest as well due to the
potential information they can bring to committees – they represent a relatively credible
source of information. While think tanks and universities can certainly be politically moti-
vated or biased, when compared to other witness types (such as witnesses from corporations
or trade associations), the research from think tanks and universities hold relatively more
scientific weight due to their connections to academic research. This result, then, points to
congressional committees seeking out and obtaining relatively more information from think
tanks and universities at the development stages of the policy-making process rather than at
later stages when a specific bill is at hand. This holds true for witnesses from corporations as
25
well. Committees also tend to seek out information from corporations more during hearings
without bills than during hearings with bills attached.
The opposite, however, is true for witnesses from labor unions, trade associations, and
membership associations. Witnesses from these mass-based groups are more likely to be
invited and testify during hearings with bills attached, compared to hearings without bills
(positive significant coefficients in Figure 5). This suggests that once committees are further
along in the policy-making process and are deliberating a specific bill, they are more inter-
ested in requesting information from witnesses who represent those who will be impacted by
the legislation.
5.2 Inter-Branch Relations and Witness Invitations
We investigate the effects of inter-branch relations on who committees turn to for information
by examining how witness invitation patterns differ during periods of divided government
(when the majority party in the House is different from the party of the president) compared
to periods of unified government. We use the following regression and ordinary least squares
estimation:
Yhict = βHearing Characteristicshict + γCommittee Characteristicsct+
δCongress Characteristicst + αi + αc + αp + εhict
where the subscripts indicate hearing h, issue i, committee c, congress t, and president p.
Congress Characteristics includes Divided Government and Democratic Majority. The main
explanatory variable Divided Government equals 1 when the majority party in the House is
different from the party of the president and equals 0 otherwise. Democratic Majority equals
1 when the Democratic Party is in majority of the House and equals 0 otherwise (when the
Republican Party is in the majority). Both Divided Government and Democratic Majority
are at the congress-level; in order to estimate the effects of these variables that vary by
congress, we include president fixed effects (αp). Committee-level and hearing-level control
26
variables (i.e. the number of witnesses in a hearing) are included as controls, as previously.
The outcome variable Yhijp will measure the percentage of witnesses in a given hearing that
are from an affiliation type.
We present the coefficient of estimating the effect of Divided Government variable on
a selected set of outcome variables in Figure 6. The full results, including outcomes of all
affiliation types and all control variables, are presented in Appendix Table A3.
Figure 6 – The Effect of Divided Government on Witness Invitations
−4.00
−2.00
0.00
2.00
4.00
No.
Witn
ess
Diver
sity
Burea
ucra
t
Con
gres
sion
al
Res
earc
h
Notes: Each plot indicates the regression coefficient for each outcome measure (x-axis). Y-axis shows
the regression coefficients; “No. Witness” is the number of witnesses, “Diversity” is the Herfindahl
index, and the others are the percentage shares of witnesses. The plots are presented with 95%
confidence interval.
Our analysis do not show a relationship between divided government and the number of
witnesses invited to testify at a hearing or the diversity of witness types. However, our results
do show that there is a negative, statistically significant effect of divided government on the
percentage of witnesses that a committee invites from the bureaucracy, compared to periods
of unified government. This lends support to the perspective that during divided government,
the majority party in Congress differs from the party in charge of the executive branch,
27
and committees (controlled by the majority party in Congress) are faced with the choice
of whether to invite bureaucratic witnesses that may represent the views of the opposing
party. Specifically, our results show that divided government is associated with a decrease
of 2.6 percentage points in the percentage of witnesses who are bureaucrats, a magnitude
which represents 7.5% of the mean percentage of bureaucrats who testify before committees.
The direction of this finding is of particular note and holds important implications for the
information that committees search for and receive during periods of divided government,
as our previous results show that bureaucrats are the types of witnesses that, on average,
provide relatively higher amounts of analytical information in their testimonies compared to
other types of witnesses.
While committees may invite lower rates of bureaucrats to testify before them during
periods of divided government, committees compensate for this by inviting higher rates
of witnesses from two types in particular. The coefficient plots for “Congressional” and
“Research” variables in Figure 6 show that there is a positive, statistically significant effect
of divided government on the percentage of witnesses that a committee invites from think
tanks and universities, as well as on the percentage of witnesses that come internally from
Congress. Divided government is associated with an increase of two percentage points in
the percentage of witnesses from think tanks and universities – as the mean percentage of
witnesses from this type who appear in hearings is 9.3%, this two percentage point increase
represents just over 20% of the mean percentage of witnesses of this type. Likewise, divided
government is associated with an increase of around one percentage point in the percentage
of witnesses that come internally from Congress, an effect magnitude which represents 12.5%
of the mean percentage of witnesses of that type who appear in hearings.
Additionally, we examine further variation into the effect of divided government on bu-
reaucrats as witnesses. We investigate whether a committee’s strategic decision to invite
bureaucrats as witnesses in congressional hearings also varies by the president’s issue pri-
orities. During the divided government, when committees hold hearings on issues that the
28
president prioritizes, the committee chair may be less likely to invite bureaucrats who would
represent the viewpoints of the executive branch. To measure the president’s issue priority,
we use data from Comparative Agenda Project’s State of the Union Speeches dataset, follow-
ing existing work (e.g., Krause and O’Connell 2016; Ballard and Curry Forthcoming). This
dataset provides issue information for each statement made during the president’s speeches.
We aggregate the number of issues by Congress and assign a decile for each issue area to
identify the relative issue priorities of the presidents. Then, we merge this information to
our hearings dataset in order to determine whether a hearing was held on an issue prioritized
by the president.
Figure 7 presents the results.16 High salient issues refer to the issues that are placed
in top 50% and low salient issues refer to the issue that are placed in the bottom 50% in
terms of the frequency of the State of the Union addresses in each Congress by the president.
When committees hold hearings on issues that the president does not prioritize, there is little
difference in terms of the frequency of inviting bureaucrats as witnesses between periods of
unified and divided government. However, when committees hold hearings on issues that the
president prioritizes (“High Salient Issues”), there is a clear diverging pattern: committees
invite relatively more bureaucrats into hearings when the majority party in the House and
the White House is the same but invite relatively fewer bureaucrats as witnesses when there
is divided control.
Overall, these findings suggest that during divided government, committees turn rela-
tively less to bureaucrats for information, and instead turn relatively more to think tanks,
universities, and internal congressional sources for information. The partisan divide between
the House and the executive branch, therefore, may not just result in partisan obstacles for
the congressional majority in getting their legislation signed into law, as commonly under-
16Table A4 in the Appendix presents the regression results and Figure 7 visualizes the results in column (3).The reference category is a hearing on low salient issues under unified government.
29
Figure 7 – The Effect of Divided Government on Inviting Bureaucrats as Witnesses By Presi-dential Issue Priorities
Low Salient Issues High Salient Issues
−4.00
−2.00
0.00
2.00
4.00
Change in B
ure
aucra
t as W
itness (
%)
Unifie
d
Divided
Unifie
d
Divided
Notes: Plots indicate the changes in the percent of witnesses who are bureaucrats during uni-
fied/divided government, by the president’s issue priorities. The plots are presented with 95%
confidence interval.
stood, but also holds implications for who provides more (or less) information that Congress
emphasizes and chooses to reveal that they consider during policy-making.17
5.3 Congressional Capacity and Witness Invitations
To investigate how the elimination of the OTA in 1995 affected the witness invitation patterns
of committees that depended on the OTA, we leverage the fact that committees differed in
their reliance on internal information. When analyzing the number of reports that congres-
sional committees requested from the OTA, there is substantial variation across committees.
For example, from 1990-1995 (the period for which report request data is available), the
17We also examine whether the party in control in the House is associated with witness invitation patterns.As Table A3 shows, having a Democratic majority in the House does not affect the number of witnesses orthe diversity of witnesses invited, and does not affect the invitation patterns of bureaucrats, congressional,or witnesses from think tanks or universities. However, a Democratic majority is associated with anincrease in the percentage of witnesses from labor unions, and a decrease in the percentage of witnessesfrom trade associations – supporting the close relationship often ascribed to the Democratic party andlabor (Schlozman 2015).
30
House Committee on Small Business requested only one report from the OTA, while the
Energy and Commerce and Science, Space, and Technology committees requested 55 reports
from the OTA.18 Certain committees, such as these latter two committees, demonstrate a
particular reliance on internal information, compared to other committees who hardly made
any use of the OTA and thus do not primarily rely on internally produced information.
Thus, we assign Energy and Commerce and Science, Space, and Technology as the group
impacted by the treatment – the committees who would be affected by the elimination of the
OTA. We estimate the following difference-in-differences model to examine whether witness
invitation patterns exhibit distinctive patterns in the treated committees compared to the
control group of committees that do not primarily rely on internal information:
Yhict = βTreatedc +6∑
s=1
γsCongress100+s +6∑
s=1
δt(Treatedc ·Congress100+s) +ρXhict +αi + εhict
In this equation, Yhcit indicates the outcome measures for witness characteristics at the
hearing level (for hearing h, issue i, committee c, in congress t). Treated indicates the two
House committees that had a strong reliance on internal information: the House Energy
and Commerce Committee and the House Committee on Science, Space, and Technology.
The variable Congress captures the lead time periods from the 100th Congress (1987-1988),
which is the reference congress. The main variable of interest is δt, which indicates whether
there were any significant differences in the witness invitation patterns between the treated
and control groups before and after the reform in the 104th Congress. Xhict include other
hearing-level control variables. We include an issue fixed effect (αi), and standard errors are
clustered at the committee level.
Figure 8 presents the results for two outcomes: (1) the number of witnesses testifying
at the hearing and (2) the percent of witnesses from think tanks and universities.19 In
the figures, the reference Congress is the 100th Congress; the plots cover the time-trends
18Figure A9 in the Appendix presents the distribution of the OTA assessment request by House committees.
19The regression results are presented in Appendix Tables A5 and A6.
31
from the 101st Congress to the 106th Congress, a period which covers three terms before
and three terms after the 1995 reform. There is no pre-trend in terms of the number of
witnesses invited and the witnesses from think tanks and universities before 1995. However,
after the reform, there was a clear the decline in the number of witnesses in the treated
committees that heavily relied on the support from the OTA, though the pattern disappears
in the subsequent Congresses. The decline in the number of research-based witnesses in the
treated group right after the reform was more substantial, and the pattern continues in the
subsequent Congresses. Given that the average percentage of witnesses who were research-
based witnesses before the reform was 7.3%, the coefficients presented in Figure 8 suggest
that there was at least a 24% drop in the invitation of research-based witnesses after the
OTA elimination.
Figure 8 – Elimination of the OTA on Witness Invitation Patterns
−3.00
−2.00
−1.00
0.00
1.00
2.00
Diffe
rence, tr
eate
d−
contr
ol com
mitte
es
101 102 103 104 105 106Congress
Number of Witness
(a) Number of Witness
−6.00
−4.00
−2.00
0.00
2.00
4.00
Diffe
rence, tr
eate
d−
contr
ol com
mitte
es
101 102 103 104 105 106Congress
Invitation of Research Witness
(b) Thinktank/Research Witness (%)
Notes: The reference Congress is 100th. Reform took place in the 104th Congress. The plots are presentedwith 95% confidence intervals.
These decreases confirms the expectation from Congressional Capacity Hypothesis 3B :
Committees that relied more on the OTA will invite relatively fewer witnesses from think
tanks and research organizations after the elimination of the OTA. This is contrary to the
expectation in the opposing hypothesis Congressional Capacity Hypothesis 3A, which was
of the view that committees who had relied heavily on the OTA may, in fact, be expected
to increase their efforts in inviting external witnesses, especially witnesses who can provide
32
technical and analytical information, in order to compensate for the loss of internal infor-
mation that had been provided from the OTA. However, a simultaneous cut in number of
committee staff across all committees in 1995 – those who play a key role in the selection,
invitation, and preparation process of witnesses, especially for technical and scientific wit-
nesses – is possibly one reason why committees who had relied on the OTA were unable to
fill the void created by the elimination of the OTA. A committee’s own staff would already
be a weaker substitute to OTA staffers – the chair of the U.S. House Committee on Science,
Space, and Technology clearly stated in 2019 that “committee staff are not a replacement
for OTA” (Johnson, 2019) – but even so, committee staff was cut as well.
Taken together, internal congressional support agencies and congressional committee staff
largely arm committees with the ability to gather and process information – these two types
of internal capacity can be characterized as “tools” that committees possess to conduct
information searches. The 1995 reform eliminated one internal source of information, the
OTA, for the specific committees that relied on this internal information. Our difference-in-
differences results reveal that these committees suffered a drop in the number of witnesses,
especially the number of research-based witnesses, as a result of the OTA elimination, and
likely could not compensate for this loss of information because of the commensurate cut to
committee staff across Congress.
6 Conclusion
In this paper, we have examined the information flow between Congress and witnesses from
external groups using a new, comprehensive dataset on committee hearings, witnesses, and
witness testimonies from 1960-2018. Overall, we use witness testimony to examine an im-
portant avenue of information-seeking behavior of congressional committees. Using 72,871
hearings and 757,161 witnesses who testified in Congress across this time period, our find-
ings show how different types of witnesses provide different levels of analytical information in
33
their testimonies and how institutional settings can affect who committees invite to provide
information. We highlight our main results and suggest extensions for future work below, to
further emphasize how our data can be of value to any scholars and policy-makers interested
in the information flow between Congress and external groups.
Our results illustrate how committees seek out different types of witnesses based on
institutional settings. For one, our results reveal that committees turn to different types of
witnesses and different types of groups based on committee intent: if they are exploring a
legislative issue and thus likely to be learning information about a potential area for future
legislation, or if they are actively considering a specific bill and thus likely to be gathering
information to craft a message surrounding the bill. This, in turn, suggests that different
groups may have different kinds of opportunities for influence through information provision
during different stages of committee politics; extensions that closely examine this and the
implications of such opportunities may be of further interest to scholars of interest group
politics.
In addition, we find that committees react to the partisan setting of divided government
by inviting lower percentages of bureaucrats to testify. This link between divided government
and lower invitation rates of bureaucrats not only has implications for the information that
committees receive, as bureaucrats have been shown to provide high levels of analytical
information in their testimonies, but also points to how committees may be choosing to
respond to partisan considerations over informational considerations. This motivates possible
future work that examines the extent to which committees may be behaving strategically with
bureaucrats. More broadly, as bureaucrats are one of the most common types of witnesses
to appear before committees, as shown in our data, using bureaucrat testimonies may be
particularly promising for future work on the inter-branch sharing of information between
congressional committees and the executive agencies.
On a final note, by scaling the amount of analytical information present in witness tes-
timonies to provide a descriptive look into how witness information can differ, we show how
34
witnesses from executive agencies, think tanks, universities, and trade associations provide
the relative highest proportions of analytical information in their testimonies. While we
focus on the level of analytical information in witness testimonies due to the existing litera-
tures focus on such information, various other ways of characterizing the content of witness
testimonies may be of further interest. For instance, the quality of information and the use
of scientific evidence in policy-making have been salient for governments and the American
public, especially in regards to current issues such as climate change, cybersecurity, and
pandemics. As Congress has made use of congressional hearings and witnesses to address
and respond to current issues such as these, further inquiry into the presentation and use
of scientific evidence in witness testimonies can enrich scholars understanding of the role of
research and scientific evidence in shaping public policy in the U.S.
35
References
Arnold, Douglas. 1990. The Logic of Congressional Action. Yale University Press.
Austen-Smith, David. 1993. “Information and Influence: Lobbying for Agendas and Votes.”American Journal of Political Science 37 (3): 799-833.
Ballard, Andrew, and James Curry. Forthcoming. “Minority Party Capacity in Congress.”American Political Science Review.
Banks, Jeffrey S., and Barry R. Weingast. 1992. “The Political Control of Bureaucraciesunder Asymmetric Information.” American Journal of Political Science 36 (2): 509-524.
Baumgartner, Frank, and Beth Leech. 1998. Basic Interests: The Importance of Groups inPolitics and in Political Science. Princeton University Press.
Baumgartner, Frank, and Bryan Jones. 2015. The Politics of Information: Problem Defini-tion and the Course of Public Policy in America. University of Chicago Press.
Baumgartner, Frank R., and Bryan D. Jones. 1993. Agendas and Instability in AmericanPolitics. University of Chicago Press.
Bendor, Jonathan, Serge Taylor, and Roland Van Gaalen. 1987. “Politicians, Bureaucrats,and Asymmetric Information.” American Journal of Political Science 31 (4): 796-828.
Bonica, Adam. 2016. “Database on Ideology, Money in Politics, and Elections: Public Version2.0.” Stanford, CA: Stanford University Libraries (https://data.stanford.edu/dime).
Bradley, R. 1980. “Motivations in Legislative Information Use.” Legislative Studies Quarterly5 (3): 393-405.
Burgat, Casey, and Charles Hunt. 2020. “How Committee Staffers Clear the Runway forLegislative Action in Congress.” In Congress Overwhelmed: The Decline in CongressionalCapacity and Prospects for Reform, ed. Timothy LaPira, Lee Drutman, and Kevin Kosar.University of Chicago Press.
Burstein, Paul. 1999. Discrimination, Jobs, and Politics. University of Chicago Press.
Carlson, David, and Jacob M. Montgomery. 2017. “A Pairwise Comparison Framework forFast, Flexible, and Reliable Human Coding of Political Texts.” American Political ScienceReview 111 (4): 835-843.
Curry, James. 2015. Legislating in the Dark: Information and Power in the House of Repre-sentatives. Chicago: University of Chicago Press.
Davis, Christopher. 2015. “House Committee Hearings: Arranging Witnesses.” Congres-sional Research Service.
Deering, Christopher, and Steven Smith. 1997. Committees in Congress. 3rd ed. CQ Press.
36
Epstein, David, and Sharyn O’Halloran. 1999. Delegating Powers: A Transactional CostsPolitics Approach to Politics under the Separation of Power. Cambridge University Press.
Esterling, Kevin. 2004. The Political Economy of Expertise: Information and Efficiency.Ann Arbor: The University of Michigan Press.
Esterling, Kevin. 2007. “Buying Expertise: Campaign Contributions and Attention to PolicyAnalysis in Congressional Committees.” American Political Science Review 101 (1): 93-109.
Flemming, Roy, Michael MacLeod, and Jeffrey Talbert. 1998. “Witnesses at the Confirma-tions? The Appearance of Organized Interests at Senate Hearings of Federal JudicialAppointments, 1945-1992.” Political Research Quarterly 51 (3): 617-631.
Gailmard, Sean, and John Patty. 2012. “Formal Models of Bureaucracy.” Annual Review ofPolitical Science 15: 353-377.
Hansen, John Mark. 1991. Gaining Access: Congress and the Farm Lobby, 1919-1981. Uni-versity of Chicago Press.
Heitshusen, Valerie. 2017. “Senate Committee Hearings: Arranging Witnesses.” Congres-sional Research Service.
Huber, John, Charles Shipan, and Madelaine Pfahler. 2001. “Legislative and Statutory Con-trol of Bureaucracy.” American Journal of Political Science 45 (2): 330-345.
Johnson, Eddie Bernice. 2019. “Experts Needed: Options for Improved Science and Tech-nology Advice for Congress.” U.S. House Committee on Science, Space, & Technology.
Kim, In Song. 2018. “LobbyView: Firm-level Lobbying & Congressional Bills Database.”MIT Working Paper (http://web.mit.edu/insong/www/pdf/lobbyview.pdf).
Kingdon, John. 1981. Congressmens’ Voting Decisions. Harper & Row.
Kollman, Ken. 1997. “Inviting Friends to Lobby: Interest Groups, Ideologial Bias, andCongressional Committees.” American Journal of Political Science 41 (2): 519-544.
Kosar, Kevin. 2020. “Legislative Branch Support Agencies: What They Are, What They Do,and Their Uneasy Position in Our System of Government.” In Congress Overwhelmed: TheDecline in Congressional Capacity and Prospects for Reform, ed. Lee Drutman, TimothyLaPira, and Kevin Kosar. University of Chicago Press.
Krause, George, and Anne Josephe O’Connell. 2016. “Experiential Learning and PresidentialManagement of the U.S Federal Bureaucracy: Logic and Evidence from Agency LeadershipAppointment.” American Journal of Political Science 60 (4): 914-931.
Krehbiel, Keith. 1991. Information and Legislative Organization. University of MichiganPress.
37
Kriner, Douglas, and Eric Schickler. 2016. Investigating the President. Princeton UniversityPress.
Lewis, David. 2003. President and the Politics of Agency Design: Political Insulation in theUnited States Government Bureaucracy, 1946-1997. Stanford University Press.
Leyden, Kevin M. 1995. “Interest Group Resources and Testimony at Congressional Hear-ings.” Legislative Studies Quarterly 20 (3): 431-439.
Lohmann, Susanne. 1995. “Information, Access, and Contributions.” Public Choice 85 (3/4):267–284.
McGrath, Robert J. 2013. “Congressional Oversight Hearings and Policy Control.” Legisla-tive Studies Quarterly 38 (3): 349-376.
Oleszek, Walter J. 1989. Congressional Procedures and the Policy Process. CQ Press.
Park, Ju Yeon. 2021. “When Do Politicians Grandstand? Measuring Message Politics inCommittee Hearings.” Journal of Politics 83 (1): 214-228.
Patty, John, and Ian Turner. 2021. “Ex Post Review and Expert Policymaking: When DoesOversight Reduce Accountability?” Journal of Politics 83 (1): 23-39.
Schlozman, Daniel. 2015. When Movements Anchor Parties: Electoral Alignments in Amer-ican History. Princeton University Press.
Schlozman, Kay, Philip Jones, Hye Young You, Tracy Burch, Sidney Verba, and HenryBrady. 2015. “Organizations and the Democratic Representation of Interests: What DoesIt Mean When Those Organizations Have No Members?” Perspectives on Politics 13 (4):1017-1029.
Schnakenberg, Keith. 2017. “Informational Lobbying and Legislative Voting.” AmericanJournal of Political Science 61 (1): 129-145.
Stewart III, Charles, and Jonathan Woon. 2017. “Congressional Commitee Assignments,103rd to 114th Congresses, 1993-2017.”.
Tudor, Grant, and Justin Warner. 2019. “Congress Should Revive the Officeof Technology Assessment. Here’s How To Do It.” Brookings Institute Decem-ber 18 (https://www.brookings.edu/blog/fixgov/2019/12/18/congress-should-revive-the-office-of-technology-assessment-heres-how-to-do-it/).
Turner, Ian. 2019. “Political Agency, Oversight, and Bias: The Instrumental Value of Politi-cized Policymaking.” Journal of Law, Economics, & Organization 35 (3): 544-478.
Volden, Craig, and Alan Wiseman. 2014. Legislative Effectiveness in the United StatesCongress. Cambridge University Press.
Yackee, Jason Webb, and Susan Webb Yackee. 2006. “A Bias Toward Business? AssessingInterest Group Influence on the U.S. Bureaucracy.” Journal of Politics 68 (1): 128-139.
38
Supporting Information for
How Are Politicians Informed? Witness Testimonyand Information Provision in Congress
A Additional Descriptive Statistics on Witness Ap-
pearances
A.1 Number and Composition of Witnesses
Figure A1 breaks down the number of witnesses who testify by committee in the House,across time. Immediately, it is clear that there are some House committees – Appropria-tions, Ways and Means, and Commerce – who have historically invited more witnesses thanother committees. Committees focused on procedural or internal matters, such as Rules,House Administration, and Standards of Official Conduct, have historically called the lowestnumber of witnesses.
Figure A2 is similar to Figure A1 except for the Senate. Among the Senate committees,we see that committees with the highest number of witnesses are Appropriations, Interiorand Insular Affairs, and Labor and Public Welfare. Rules and Administration, similar toits counterpart in the House, is one of the committees with the lowest number of witnesses,though is joined by Veterans’ Affairs, Budget, and Foreign Relations. Of note is the factthat Foreign Relations in the Senate and its counterpart, Foreign Affairs in the House, bothhave low numbers of witnesses compared to the other committees.
Figure A3 is the Senate version of Figure 3 in the main text, and shows the averagecomposition of witness affiliations by committee in the House.
Figure A4 plots the composition of witness types (grouped by parent category for illus-trative purposes) called by each party when they are in the majority party in each chamber.The top bar in each panel shows the percentages of witnesses called of each category whenthe Republicans are in the majority (and hold all committee chairs) in that chamber. Thebottom bar in each panel shows the percentages of witnesses called of each category whenthe Democrats are in the majority (and hold all committee chairs) in that chamber.
A1
Figure A1 – Witnesses in House Standing Committees Across Time
Standards of Official Conduct Veterans' Affairs Ways and Means
Post Office and Civil Service Public Works Rules Science and Astronautics Small Business
Interior and Insular Affairs Internal Security Interstate and Foreign Commerce Judiciary Merchant Marine and Fisheries
District of Columbia Education and Labor Foreign Affairs Government Operations House Administration
Agriculture Appropriations Armed Services Banking and Currency Budget
1960 1980 2000 2020 1960 1980 2000 2020 1960 1980 2000 2020
1960 1980 2000 2020 1960 1980 2000 2020 1960 1980 2000 2020 1960 1980 2000 2020 1960 1980 2000 2020
1960 1980 2000 2020 1960 1980 2000 2020 1960 1980 2000 2020 1960 1980 2000 2020 1960 1980 2000 2020
1960 1980 2000 2020 1960 1980 2000 2020 1960 1980 2000 2020 1960 1980 2000 2020 1960 1980 2000 2020
1960 1980 2000 2020 1960 1980 2000 2020 1960 1980 2000 2020 1960 1980 2000 2020 1960 1980 2000 2020
0
1000
2000
3000
0
1000
2000
3000
0
1000
2000
3000
0
1000
2000
3000
0
1000
2000
3000
0
1000
2000
3000
0
1000
2000
3000
0
1000
2000
3000
0
1000
2000
3000
0
1000
2000
3000
0
1000
2000
3000
0
1000
2000
3000
0
1000
2000
3000
0
1000
2000
3000
0
1000
2000
3000
0
1000
2000
3000
0
1000
2000
3000
0
1000
2000
3000
0
1000
2000
3000
0
1000
2000
3000
0
1000
2000
3000
0
1000
2000
3000
0
1000
2000
3000
Year
Num
ber
of W
itnes
ses
Witnesses in House Standing Committees Across Time
A2
Figure A2 – Witnesses in Senate Standing Committees Across Time
Public Works Rules and Administration Veterans' Affairs
Government Operations Interior and Insular Affairs Judiciary Labor and Public Welfare Post Office and Civil Service
Budget Commerce District of Columbia Finance Foreign Relations
Aeronautical and Space Sciences Agriculture and Forestry Appropriations Armed Services Banking, Housing, and Urban Affairs
1960 1980 2000 2020 1960 1980 2000 2020 1960 1980 2000 2020
1960 1980 2000 2020 1960 1980 2000 2020 1960 1980 2000 2020 1960 1980 2000 2020 1960 1980 2000 2020
1960 1980 2000 2020 1960 1980 2000 2020 1960 1980 2000 2020 1960 1980 2000 2020 1960 1980 2000 2020
1960 1980 2000 2020 1960 1980 2000 2020 1960 1980 2000 2020 1960 1980 2000 2020 1960 1980 2000 2020
0
500
1000
1500
2000
0
500
1000
1500
2000
0
500
1000
1500
2000
0
500
1000
1500
2000
0
500
1000
1500
2000
0
500
1000
1500
2000
0
500
1000
1500
2000
0
500
1000
1500
2000
0
500
1000
1500
2000
0
500
1000
1500
2000
0
500
1000
1500
2000
0
500
1000
1500
2000
0
500
1000
1500
2000
0
500
1000
1500
2000
0
500
1000
1500
2000
0
500
1000
1500
2000
0
500
1000
1500
2000
0
500
1000
1500
2000
Year
Num
ber
of W
itnes
ses
Witnesses in Senate Standing Committees Across Time
A3
Figure A3 – Witness Affiliations By Senate Committee
0%
25%
50%
75%
100%
Armed
Ser
vices
Fore
ign R
elatio
ns
Appro
priat
ions
Homela
nd S
ecur
ity a
nd G
over
nmen
tal A
ffairs
Vete
rans
' Affa
irs
Bankin
g, H
ousin
g, a
nd U
rban
Affa
irs
Comm
erce
, Scie
nce,
and
Tran
spor
tatio
n
Budge
t
Small
Bus
iness
and
Ent
repr
eneu
rship
Enviro
nmen
t and
Pub
lic W
orks
Financ
e
Energ
y and
Nat
ural
Resou
rces
Judic
iary
Agricu
lture
, Nut
rition
, and
For
estry
Health
, Edu
catio
n, L
abor
, and
Pen
sions
Rules a
nd A
dmini
strat
ion
Committee
Com
posi
tion
BureaucratBusinessMembership AssocResearchCongressionalLocal GovNonprofitLaborOther
Senate
A4
Figure A4 – Witness Affiliations by Majority Party
Democrat
Republican
0% 25% 50% 75% 100%
Average Composition in a Congress
Par
ty in
Maj
ority Bureaucrat
BusinessMembership AssocResearchCongressionalLocal GovNonprofitLaborOther
(A) House
Democrat
Republican
0% 25% 50% 75% 100%
Average Composition in a Congress
Par
ty in
Maj
ority Bureaucrat
BusinessMembership AssocResearchCongressionalLocal GovNonprofitLaborOther
(B) Senate
Notes: In each panel, the top bar presents the percentages of witnesses of each affiliation category
called in that chamber when the Republicans are the majority party in that chamber. The bottom
bars present the percentages of witnesses of each affiliation category called in that chamber when
the Democrats are the majority party in that chamber.
A5
A.2 Witnesses by Issue Area
Figure A5 – Witness Affiliations by Issue Area
0%
25%
50%
75%
100%
Agricu
lture
Civil R
ights
Cultur
e
Defens
e
Domes
tic C
omm
erce
Educa
tion
Energ
y
Enviro
nmen
t
Fore
ign Tr
ade
Gover
nmen
t Ope
ratio
ns
Health
Housin
g
Imm
igrat
ion
Inte
rnat
ional
Affairs
Labo
r
Law a
nd C
rime
Mac
roec
onom
ics
Public
Lan
ds
Social
Welf
are
Tech
nolog
y
Trans
porta
tion
NA
Issue
Com
posi
tion
AssociationBureaucratBusinessCongressionalLaborLocal GovNonprofitOtherResearch
House
0%
25%
50%
75%
100%
Agricu
lture
Civil R
ights
Cultur
e
Defens
e
Domes
tic C
omm
erce
Educa
tion
Energ
y
Enviro
nmen
t
Fore
ign Tr
ade
Gover
nmen
t Ope
ratio
ns
Health
Housin
g
Imm
igrat
ion
Inte
rnat
ional
Affairs
Labo
r
Law a
nd C
rime
Mac
roec
onom
ics
Public
Lan
ds
Social
Welf
are
Tech
nolog
y
Trans
porta
tion
NA
Issue
Com
posi
tion
AssociationBureaucratBusinessCongressionalLaborLocal GovNonprofitOtherResearch
Senate
A6
A.3 Witness Types Across Time
Figure A6 – Number of Witnesses by Type: House
0
1000
2000
3000
4000
5000
6000
7000
1960 1970 1980 1990 2000 2010Year
Num
ber
of W
itnes
ses
AGRICULTUREASSOCIATIONBUREAUCRATCITIZEN
CONGRESSIONALCORPORATIONEDUCATIONHEALTHCARE
INTERNATIONALJUDICIALLABORLAWYERS&LOBBYISTS
NATIVE AMERICANNONPROFITRELIGIOUSSTATE&LOCAL GOV
THINKTANK/RESEARCHTRADE ASSOCIATION
A7
Figure A7 – Number of Witnesses by Type: Senate
0
1000
2000
3000
4000
5000
6000
7000
1960 1970 1980 1990 2000 2010Year
Num
ber
of W
itnes
ses
AGRICULTUREASSOCIATIONBUREAUCRATCITIZEN
CONGRESSIONALCORPORATIONEDUCATIONHEALTHCARE
INTERNATIONALJUDICIALLABORLAWYERS&LOBBYISTS
NATIVE AMERICANNONPROFITRELIGIOUSSTATE&LOCAL GOV
THINKTANK/RESEARCHTRADE ASSOCIATION
A8
B Measuring Analytical Information in Witness Testi-
monies
B.1 Keywords
The keywords that potentially cue that a testimony may contain some analytical informa-tion were chosen from three sources. First, we refer to the grandstanding score introduced inPark (2021) which assigns a continuous score to committee members’ statements to measurepolitical messaging activities in congressional hearings from the 105th to 114th Congresses.As a side-product of the score, members’ statements scoring low are featured largely byeither procedural statements or information-seeking statements. From the list of 200 mostfrequent word stems in the statements scoring the lower quartile of the score, we selected74 word stems that were deemed relevant to bills (e.g., bill, law and legisl), sources of in-formation (e.g. inform, letter, record and report), research (e.g., author, data, estim andstudi), statistics (e.g., percent, rank and rate), logical relationship (e.g. relat, associ anddiffer), cost-benefit calculation (e.g., benefit, budget, cost and dollar), policy consequences(e.g., change, effect, impact and increase), and deliberation (e.g., discuss, possibl, and re-view). Then, we added one more word stem and two special characters: “statist”, “%” and“$”. These word categories can be considered constituting a typical policy-making processwhich includes collecting information and data, analyzing them, assessing cost, benefit andpossible consequences of policy alternatives, and finally deliberating and making decision onthe choice of the alternatives.
Second, we additionally collected words that are related to cognitive orientation from the“Harvard IV-4” dictionary. Specifically, we chose 32 words in the following sub-categories:”know” (e.g., analyt, calcul and correl), ”causal” (e.g., caus, consequ and odd), ”com-pare” (e.g., less, higher and better) and ”quan” (e.g., approx, averg and disproportion)and stemmed the words for the analysis.
Third, to complement the list, we identify 28 more word stems that are relevant toanalytical information but not in the list of words described above (e.g., diagnosi, survey,examin, investig and measure) or the words that have similar meaning with that of the wordsin this list but not included in the list (e.g., percentag is similar to ”percent”; contrast issimilar to ”differ”; result is similar to ”consequ”). In total, we use 134 keyword stems forthis study. The full list of the keywords is in the appendix.
B.2 The List of the 134 Keyword Stems
$, %, address, analit, analysi, analyt, answer, approxim, assess, associ, author, averag,awar, benefit, better, bill, budget, calcul, case, caus, chang, classif, classifi, comment, com-par, comparison, consequ, consid, content, contrast, contribut, correct, correl, cost, criteria,data, decid, decis, decreas, degre, determin, determinist, diagnosi, diagnost, differ, discuss,disproportion, dollar, effect, empir, equival, estim, evid, examin, explain, fact, factor, fea-sibl, fund, higher, impact, implaus, imposs, improv, increas, indic, influenc, inform, interest,investig, laboratori, law, legisl, less, letter, level, list, lower, mean, measur, necessari, need,number, object, odd, percent, percentag, plan, plausibl, point, polici, possibl, predict, prob-abl, process, product, project, propos, rais, rank, rate, reason, recommend, record, reduc,
A9
refer, relat, report, requir, research, respond, respons, result, review, rise, risk, scienc, scien-tif, solut, solv, specif, standard, statement, statist, studi, substanti, survey, technolog, test,testifi, understand, unit, wors, yield
B.3 The Most and Least Analytical Testimony
B.3.1 With the length limit to include 50 to 150 words
The most analytical statements
1. “When projects are authorized, when there is a Chief’s Report and the Congressauthorizes a project, the economic analysis that is done on that calculates a benefitto cost ratio. And that benefit to cost ratio is based on a 3.125 discount rate. Whenthe Office of Management and Budget evaluates projects for funding, including in thePresident’s budget, that benefit to cost ratio is evaluated at a 7-percent discount rate.So the budgeting discount rate is different from the authorization discount rate that’sused.”
2. “We found that the differences are primarily–and this is a big amount of–the biggestchunk was in the estimate of labor costs associated with the subcontractors. There werecosts also associated–of $1.2 billions–associated with engine cost that was a differencein the estimate; also $1 billion in terms of the production cost reduction plans, andalso $800 million difference in terms of what the Air Force’s plans for–relating toproductivity investments.”
3. “In terms of offsetting the costs and benefits, we did offset those costs, so the benefitsare reduced by the amount of those costs in terms of attributing–and that’s in thecost/benefit analysis, but in analyzing the costs and in analyzing the benefits, we didreduce the benefits by those costs.”
A10
The least analytical statements
1. “Now, the access through public lands is, again, a heated debate. The President justdrew an Executive Order declaring much of the border area and New Mexico as amonument, wilderness, whatever. They are all the same. Is the Organ Pipe NationalMonument, has that still got the signs up there requesting people not to go in there,American citizens, saying you should not go in there because it is too dangerous?”
2. “I guess we mistakenly believed that it was a secret location, and the only people whoknew about it were the EOD staff from both SFPD, the FBI and the Sheriff’s Office.Unbeknownst to us, this particular individual, and I won’t say too much, but was aplumber in that area and apparently had seen the officers going into that area andperhaps followed them in.”
3. “And don’t forget by the way, sir, that we have right now–and the senator gets upsetabout this, but you have time to do this. We should do it this year. But we shouldadjust the system so that we get ready for 2017 when more money is going out thancoming in, and we can do it.”
B.3.2 Without the length limit
The most analytical statements
1. “Well, when you say higher costs, higher costs overall or higher costs—-”
2. “It would increase confidence, lower expected tax rates, and lower real interest rates.”
3. “That is correct. The President’s budget proposes a funding level of $100 million.”
The least analytical statements
1. “Thank you. I am going to ask my colleague, Mike Connor, to take that question.”
2. “Thank you very much, Mr. Souder, and your staff for helping to deal me in today.I found out about this yesterday morning, and I’m pleased to be here. I am a formercollege administrator and teacher. My name is Dean, but I was one once.”
3. “If Congress would like to do that, I would be absolutely thrilled.”
A11
B.4 The Statistical Validation Strategy for the Measurement ofAnalytical Information
This section explains how we constructed a human-coded validation measurement for the 100sample paragraphs of witness testimonies. First, we randomly selected 1000 statements thatwitnesses made and keep only the statements with more than 80 words. Then, if a statementcontains multiple paragraphs, we divide the statement by paragraph. Among the paragraphsor single-paragraph statements, we keep only those with less than 50 words or more than150 words. Second, we measure the proportion of keywords for each paragraph. Third, weconduct random block sampling to construct 100 sample paragraphs to be human-coded;we select 20 paragraphs from each of the following five blocks: 0-0.05, 0.05-0.1, 0.1-0.15,0.15-0.2, 0.2 or above. The thresholds are chosen such that they divide the range thatthe proportion of keywords in our data runs into five equidistant smaller ranges. Fourth,each of the 100 sample paragraphs are randomly matched with another paragraph to create1000 pairs. Fifth, each of the two trained student research assistants compares 500 pairsand chooses the one that sounds more analytical. To define analytical information, weborrow the definition of analytical information from Esterling (2007). That is, a paragraphis analytical if it contains verifiable, fact-based, objective or positive statement as opposed tonon-verifiable, experiential, opinion-based, subjective or normative. After collecting coders’choices, we fit a STAN model to measure the latent trait in the sample paragraphs andconstruct a continuous measurement as suggested in Carlson and Montgomery (2017).
The correlation coefficient between our measurement, the proportion of keywords, andthe human-coded score resulting from the STAN model is 0.6, which provides statistical aswell as substantive validation of our measurement. This correlation shows that they runin the same direction and this validation strategy is considered suitable for the purpose ofshowing descriptive analysis about the differences across witnesses’ affiliation types.
A12
B.5 Regression and Results
The regression equation is shown below:
Proportion of keywordssfhict = α0+β∗Hearing Characteristicsh+γ∗Committee Characteristicsct
+αf + αi + αc + αt + εsfhict
where the subscripts indicate statements s, witness affiliations f , hearings h, issue i, com-mittee c, and congress t.
In these regression models, we control for the following control variables. At the hearing-level, we control for the number of times that a witness was asked to speak in a hearing,an indicator for whether a bill was considered, the number of committee members present,the number of witnesses present in a hearing, and a subcommittee hearing indicator. Atthe committee level, we include the ideological distance between the floor median and thecommittee median based on the DW-NOMINATE score to capture how ideologically extremethe committee is as a group, the distance between Democrats and Republicans in a committeeto capture the level of polarization within a committee, the distance between the floor medianand the committee chair to measure the ideological intensity of the chair, and the averagelegislative effectiveness score of the committee members who spoke in a hearing (Volden andWiseman 2014). We also include congress fixed effects, committee fixed effects, hearing issuefixed effects (from the Policy Agendas Project), and witness affiliation fixed effects.
The results from this regression is shown in Table A1 next.
A13
Table A1 – Regression Results Analyzing Witness Testimonies
Dependent variable:
Words Keywords Keywords/Words
(1) (2) (3)
Number of Statements 66.521∗∗∗ 3.391∗∗∗ −0.0001∗∗∗
(0.331) (0.020) (0.00001)Bill −91.301∗∗∗ −3.313∗∗∗ 0.0005∗∗
(10.698) (0.637) (0.0002)Number of Members 337.180∗∗∗ 11.059∗∗∗ 0.0004
(53.204) (3.166) (0.001)Number of Witnesses −29.297∗∗∗ −1.686∗∗∗ −0.0001∗∗∗
(0.994) (0.059) (0.00002)Subcommittee Hearing −44.006∗∗∗ 0.020 0.001∗∗
(13.578) (0.808) (0.0003)Committtee Ideology −554.421∗∗∗ −11.078∗ 0.009∗∗∗
(106.764) (6.353) (0.002)Polarization of Floor −679.374∗∗∗ −38.686∗∗∗ −0.0004
(116.496) (6.932) (0.002)Chair’s Ideology −248.777∗∗∗ −12.171∗∗∗ 0.0001
(51.078) (3.040) (0.001)Avg. LES of Committee 7.279∗ 0.404 −0.00002
(4.327) (0.257) (0.0001)Constant 2, 199.623∗∗∗ 108.552∗∗∗ 0.047∗∗∗
(91.597) (5.451) (0.002)
Witness Type FE Yes Yes YesIssue FE Yes Yes YesCommittee FE Yes Yes YesCongress FE Yes Yes Yes
Observations 33,605 33,605 33,605R2 0.652 0.604 0.149Adjusted R2 0.652 0.603 0.147
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01The dependent variable in the first model is the number of words spoken; in the second,the number of keywords spoken; and in the third, the proportion of keywords in thetotal number of words spoken.
A14
Figure A8 – Number of Keywords by Witness Type
Notes: Vertical lines indicate 95% confidence interval.
A15
C Institutional Conditions and Witness Invitation
Table A2 – Hearing Characteristics and Witness Invitation Patterns
Panel A (1) (2) (3) (4) (5) (6) ( 7) (8)Outcome (%) = No.Witness Diversity Bureau Research Corp. Labor Trade Membership
Bill 2.123∗∗∗ 6.460∗∗∗ -7.605∗∗∗ -1.919∗∗∗ -1.551∗∗∗ 0.578∗∗ 1.939∗∗∗ 3.056∗∗∗
(0.314) (0.522) (0.765) (0.365) (0.385) (0.247) (0.497) (0.490)
Subcommittee -0.896 6.228∗∗∗ -5.019∗∗∗ 0.593 1.216∗∗∗ 0.0450 0.0750 0.838∗∗
(0.548) (0.736) (1.682) (0.830) (0.414) (0.173) (0.530) (0.382)
No. Comm. Members 0.0403 -0.0778 -0.0234 -0.00754 0.0100 0.0202 0.0526∗∗ 0.0917(0.0448) (0.0592) (0.0994) (0.0424) (0.0336) (0.0202) (0.0204) (0.0551)
|Floor Median-Comm. Median| -0.112 6.150 1.592 -5.113 -4.653∗∗∗ 1.176 5.060∗∗∗ 6.209∗
(4.393) (3.831) (8.129) (4.155) (1.608) (1.413) (1.688) (3.272)
|Comm. Dem-Comm. Rep| 5.022∗ -6.330∗ 6.439 4.844∗ -0.660 1.322 -0.267 -3.379(2.887) (3.351) (4.951) (2.397) (1.665) (0.986) (1.326) (2.025)
|Floor Median-Comm. Chair| 2.243 -0.194 4.686 -1.854 -0.0650 0.544 -0.212 -2.980(1.645) (3.043) (3.770) (1.616) (0.863) (0.531) (0.900) (1.908)
Number of Witness 1.045∗∗∗ -1.009∗∗∗ 0.0668∗∗ 0.119∗∗∗ 0.0285∗∗∗ 0.105∗∗∗ 0.145∗∗∗
(0.0988) (0.0909) (0.0267) (0.0229) (0.00702) (0.0205) (0.0216)
N 30994 30983 30983 30983 30983 30983 30983 30983
adj. R2 0.157 0.318 0.288 0.128 0.130 0.166 0.161 0.224Mean Outcome Var. 9.8 53.6 34.8 9.3 8.1 2.2 5.7 7.8
Panel B (9) (10) (11) (12) (13) (14) (15) (16)Outcome (%) = Agri. Cong. Judicial Local Gov. Lawyer Nonprofit Healthcare Other
Bill -0.106 6.194∗∗∗ 0.222 -1.216∗∗∗ 0.153 0.775∗∗∗ -0.0612 -0.459∗
(0.0869) (0.518) (0.185) (0.316) (0.116) (0.245) (0.102) (0.222)
Subcommittee 0.155 0.901∗∗ 0.0196 0.559 -0.0799 1.551∗∗∗ 0.181 -1.034(0.0994) (0.352) (0.0860) (0.486) (0.143) (0.402) (0.154) (1.084)
No. Comm. Members -0.0142∗ -0.0726∗ 0.00255 0.00787 -0.0149 -0.0176 -0.0162∗ -0.0185(0.00783) (0.0387) (0.00510) (0.0284) (0.00996) (0.0202) (0.00885) (0.0175)
|Floor Median-Comm. Median| 0.496 0.0771 -1.714 -3.551 -0.648 0.715 -0.578 0.932(0.633) (3.126) (1.114) (2.442) (0.903) (1.714) (0.685) (1.022)
|Comm. Dem-Comm. Rep| -0.742 -4.718∗ 0.550 -1.133 -0.259 -0.680 0.305 -1.622(0.601) (2.702) (0.659) (0.980) (0.718) (1.396) (0.573) (1.267)
|Floor Median-Comm. Chair| 0.521 -1.496 -0.173 0.485 -0.213 -0.420 0.154 1.023(0.482) (1.462) (0.224) (1.116) (0.290) (0.952) (0.410) (1.185)
Number of Witness 0.0358∗∗ 0.0946∗∗ -0.00596 0.200∗∗∗ 0.000859 0.120∗∗∗ 0.0196∗∗∗ 0.0789∗∗∗
(0.0170) (0.0351) (0.00541) (0.0283) (0.00287) (0.0172) (0.00384) (0.0107)
N 30983 30983 30983 30983 30983 30983 30983 30983
adj. R2 0.332 0.146 0.087 0.175 0.065 0.091 0.253 0.074Mean Outcome Var. 1.0 7.7 0.6 8.5 1.4 6.7 1.4 4.1
∗ p < 0.10 ∗∗ p < 0.05, ∗∗∗ p < 0.01. Congress, Committee, Issue FEs are included.
A16
Table A3 – Institutional Characteristics and Witness Invitation Patterns
Panel A (1) (2) (3) (4) (5) (6) ( 7) (8)Outcome (%) = No.Witness Diversity Bureau Cong. Research Agri. Corp. Trade
Divide Government -0.468 0.313 -2.613∗∗ 0.965∗∗ 2.151∗∗∗ 0.161∗ 0.238 -0.0239(0.340) (0.766) (0.941) (0.344) (0.691) (0.0818) (0.405) (0.272)
Democratic Majority 0.150 0.450 -1.421 -0.375 1.374∗ -0.379∗∗ 0.486∗ -1.172∗∗∗
(0.319) (1.152) (1.217) (0.438) (0.727) (0.137) (0.272) (0.341)
Bill 2.149∗∗∗ 6.422∗∗∗ -7.533∗∗∗ 6.188∗∗∗ -1.943∗∗∗ -0.106 -1.574∗∗∗ 1.926∗∗∗
(0.317) (0.540) (0.785) (0.522) (0.362) (0.0842) (0.376) (0.501)
Bubcommittee -0.909 6.131∗∗∗ -4.961∗∗∗ 0.854∗∗ 0.580 0.147 1.199∗∗∗ 0.0621(0.545) (0.732) (1.660) (0.341) (0.834) (0.0976) (0.415) (0.526)
No. Comm. Members 0.0352 -0.0166 -0.0507 -0.0765∗ 0.00658 -0.0142∗ 0.0151 0.0527∗∗
(0.0402) (0.0551) (0.0910) (0.0390) (0.0389) (0.00698) (0.0302) (0.0203)
|Floor Median-Comm. Median| -0.187 6.737 0.984 0.251 -4.310 0.369 -4.667∗∗∗ 5.078∗∗∗
(4.287) (4.797) (8.459) (3.074) (4.120) (0.532) (1.564) (1.591)
|Comm. Dem-Comm. Rep| 5.418∗ -6.674∗ 6.445 -5.356∗ 4.514∗ -0.746 -0.211 -0.0954(2.741) (3.445) (4.890) (2.614) (2.290) (0.610) (1.544) (1.330)
|Floor Median-Comm. Chair| 1.974 -0.812 4.409 -1.565 -1.564 0.475 -0.308 -0.209(1.629) (3.241) (3.682) (1.495) (1.515) (0.466) (0.978) (0.933)
Number of Witness 1.043∗∗∗ -1.005∗∗∗ 0.0965∗∗ 0.0650∗∗ 0.0355∗∗ 0.119∗∗∗ 0.105∗∗∗
(0.0987) (0.0910) (0.0351) (0.0267) (0.0169) (0.0229) (0.0205)
N 30994 30983 30983 30983 30983 30983 30983 30983
adj. R2 0.154 0.316 0.287 0.145 0.128 0.332 0.130 0.161Mean Outcome Var. 9.8 53.6 34.8 7.7 9.3 1.0 8.1 5.7
Panel B (9) (10) (11) (12) (13) (14) (15) (16)Outcome (%) = Judicial Local Gov. Lawyer Labor Nonprofit Healthcare Membership Other
Divided Government -0.0387 0.108 0.290∗ -0.410∗∗∗ -0.224 -0.214 -0.659 0.271(0.0504) (0.270) (0.152) (0.135) (0.278) (0.163) (0.396) (0.365)
Democratic Majority 0.0723 -0.287 0.361∗∗ 0.340∗∗∗ 0.572 -0.0582 -0.114 0.601(0.0993) (0.457) (0.172) (0.100) (0.531) (0.162) (0.484) (0.564)
Bill 0.223 -1.212∗∗∗ 0.152 0.584∗∗ 0.774∗∗∗ -0.0632 3.057∗∗∗ -0.475∗∗
(0.185) (0.320) (0.116) (0.245) (0.246) (0.103) (0.493) (0.215)
Subcommittee 0.0210 0.558 -0.0684 0.0518 1.577∗∗∗ 0.176 0.842∗∗ -1.041(0.0848) (0.481) (0.141) (0.173) (0.403) (0.152) (0.369) (1.075)
No. Comm. Members 0.00256 0.0176 -0.0131 0.0176 -0.0137 -0.0137 0.0861∗ -0.0162(0.00510) (0.0283) (0.00923) (0.0187) (0.0196) (0.00837) (0.0491) (0.0183)
|Floor Median-Comm. Median| -1.707 -3.758 -0.579 0.877 0.902 -0.436 6.057∗ 0.938(1.110) (2.494) (0.979) (1.284) (1.618) (0.676) (3.150) (1.043)
|Comm. Dem-Comm. Rep| 0.445 -0.644 -0.297 1.456 -0.860 -0.0215 -3.424∗ -1.205(0.632) (0.963) (0.650) (1.030) (1.409) (0.604) (1.957) (1.124)
|Floor Median-Comm. Chair| -0.112 0.596 -0.135 0.449 -0.399 0.174 -2.894 1.084(0.196) (1.182) (0.327) (0.508) (0.926) (0.406) (1.982) (1.078)
Number of Witness -0.00611 0.200∗∗∗ 0.000304 0.0286∗∗∗ 0.119∗∗∗ 0.0192∗∗∗ 0.145∗∗∗ 0.0781∗∗∗
(0.00528) (0.0284) (0.00290) (0.00723) (0.0172) (0.00383) (0.0213) (0.0107)
N 30983 30983 30983 30983 30983 30983 30983 30983
adj. R2 0.087 0.175 0.065 0.165 0.090 0.253 0.225 0.074Mean Outcome Var. 9.6 8.5 1.4 2.2 6.7 1.4 7.8 4.1
∗ p < 0.10 ∗∗ p < 0.05, ∗∗∗ p < 0.01. President, Committee, Issue FEs are included.
A17
Table A4 – Divided Government, President’s Issue Priority, and Bureaucrats as Witnesses
(1) (2) (3)Outcome = Bureaucrat as Witness (%)Divided Government -2.153∗∗ -0.292 -1.304∗
(0.796) (0.897) (0.694)
Issue Decilea 0.401∗∗
(0.169)
Divided Government × Issue Decile -0.384∗∗
(0.147)
High Salient Issueb 1.704∗∗
(0.687)
Divided Government × High Salient Issue -1.562∗∗
(0.683)
Controls 3 3 3N 31773 27270 31773adj. R2 0.275 0.277 0.275∗ p < 0.10 ∗∗ p < 0.05, ∗∗∗ p < 0.01. President and committee FEs are included.Standard errors are clustered at the committee level. Hearing- and committee-levelcontrols are included. a: President’s issue priority measure based on the State ofthe Union speeches. It ranges from 1 to 10: 1 = least frequently mentioned issue,10 = most frequently mentioned issue. b: 1 if Issue Decile ≥ 5 and 0 otherwise.
A18
Table A5 – Elimination of OTA on the Number of Invited Witness
Variable Coef. Std. Err. t-stat P-value [95% Conf. Interval]
Treated -0.0183 0.6563 -0.03 0.978 -1.3833 1.3466101th Congress 0.0208 0.2794 0.07 0.941 -0.5603 0.6020102th Congress -0.7329 0.4251 -1.72 0.099 -1.6169 0.1511103th Congress -0.9991 0.6318 -1.58 0.129 -2.3130 0.3148104th Congress 0.9819 0.5590 1.76 0.094 -0.1806 2.1444105th Congress -1.6116 0.6859 -2.35 0.029 -3.0381 -0.1851106th Congress -1.9415 0.6694 -2.9 0.009 -3.3337 -0.5493treatedX101th Congress -0.6811 0.4862 -1.4 0.176 -1.6922 0.3300treatedX102th Congress -0.0100 0.5651 -0.02 0.986 -1.1852 1.1652treatedX103th Congress -0.2014 0.8994 -0.22 0.825 -2.0718 1.6690treatedX104th Congress -2.0086 0.6082 -3.3 0.003 -3.2734 -0.7438treatedX105th Congress -0.5624 0.7880 -0.71 0.483 -2.2012 1.0764treatedX106th Congress -1.1619 0.7721 -1.5 0.147 -2.7676 0.4439Bill 2.2112 0.4534 4.88 0 1.2684 3.1541Subcommittee -1.1215 0.8328 -1.35 0.192 -2.8534 0.6104Number of Committee Member -0.0154 0.0413 -0.37 0.712 -0.1014 0.0705
Notes: Number of observation is 10,179. Prob >F = 0.0000. Adj R-squared = 0.0677. Issue fixedeffects are included. Standard errors are clustered at the committee level.
Table A6 – Elimination of OTA on the Invitation of Research Witness
Variable Coef. Std. Err. t-stat P-value [95% Conf. Interval]
Treated 3.993364 2.7699 1.4400 0.1640 -1.766887 9.753615101th Congress -0.3568618 0.4594 -0.7800 0.4460 -1.312171 0.5984473102th Congress 1.250601 0.9309 1.3400 0.1930 -0.6853599 3.186562103th Congress -0.5356855 0.8278 -0.6500 0.5250 -2.257258 1.185887104th Congress 1.045746 0.9077 1.1500 0.2620 -0.8420227 2.933514105th Congress 1.103274 0.9820 1.1200 0.2740 -0.9388484 3.145397106th Congress 0.7270946 0.9801 0.7400 0.4660 -1.31112 2.765309treatedX101th Congress 0.0767433 0.5936 0.1300 0.8980 -1.157769 1.311256treatedX102th Congress 0.4090257 1.8902 0.2200 0.8310 -3.521845 4.339896treatedX103th Congress 0.6340882 1.0286 0.6200 0.5440 -1.504983 2.773159treatedX104th Congress -4.594514 0.6603 -6.9600 0.0000 -5.967743 -3.221285treatedX105th Congress -1.748871 0.8168 -2.1400 0.0440 -3.447461 -0.0502822treatedX106th Congress -3.706158 0.9647 -3.8400 0.0010 -5.712429 -1.699888Bill -1.811747 0.5606 -3.2300 0.0040 -2.977584 -0.6459086Subcommittee -2.064094 1.5808 -1.3100 0.2060 -5.351648 1.22346Number of Committee Member 0.0176605 0.0659 0.2700 0.7910 -0.1193818 0.1547028
Notes: Number of observation is 10,172. Prob >F = 0.0000. Adj R-squared = 0.0787. Issue fixed effectsare included. Standard errors are clustered at the committee level.
A19
Figure A9 – Number of OTA Assessment Request by House Committees, 1990-1995
0 20 40 60Number of OTA Requests, 1990−1995
Science, Space and Technology
Energy and Commerce
Government Operations
Natural Resources
Foreign Affairs
Ways and Means
Appropriations
Education and Labor
Agriculture
Armed Services
Judiciary
Veterans’ Affairs
Transportation
Financial Services
Rules
Budget
Small Business
Figure A10 – Changes in the Number of Committee Staff
A20
top related